Home / Methodology
Data Governance

Methodology, in full.

Research is only as good as its provenance. This page documents every source we use, how often it updates, what we derive from it, and — importantly — what our data cannot tell you. If you intend to make decisions using our work, read this first.

Last updated 26 August 2026

1 · Data sources

We use public and commercially available sources only. We do not use non-public information, and we do not accept data from any party with a position in the markets we cover.

SourceCoversClassLicence
gold-api.comGold, silver, platinum, palladium spotreal-timePublic API
Yahoo Finance chart APICrude (WTI/Brent), natural gas, coal, wheat, corn, soybeans, coffee, cocoa, sugar, copper15-min delayedPublic endpoint
Yahoo Finance (ETF/ETN)Lithium, cobalt, nickel, rare earthsproxyPublic endpoint
Open-Meteo forecast APISoil moisture, soil temperature, rainfall, reference evapotranspiration, max temperature — 16 growing regionsdailyCC-BY 4.0
Open-Meteo archive API366 days of historical weather per regionarchiveCC-BY 4.0
IEA, USGS, World Bank, IDB, Afreximbank, Australian DISRStructural production shares and reserve data in the S&D matrixestimatePublished reports
Why proxies are labelled as proxies. Lithium, cobalt, nickel and rare earth prices are not freely available as spot benchmarks. Where we use an exchange-traded fund or note as a directional proxy, we say so on every table row. A proxy tracks sentiment and broad direction; it is not the physical price and should never be used as a settlement reference.

2 · Capture and frequency

The desk captures every three hours, continuously. Each capture writes a timestamped snapshot to our archive; nothing is overwritten, so the series is reconstructible after the fact.

  • Cadence: every 3 hours (00:00, 03:00, 06:00, 09:00, 12:00, 15:00, 18:00, 21:00 UTC)
  • Retention: permanent. Daily JSON snapshots plus an append-only CSV series
  • Resilience: each source has retry with backoff; metals have a secondary fallback feed
  • Publication gate: if fewer than 15 of 19 instruments return data, the capture is recorded but not published. We would rather show yesterday's complete picture than today's partial one

3 · Labelling standard

Every number we publish carries its class. This is deliberate and non-negotiable — a reader must be able to tell at a glance how much weight a figure can bear.

LabelMeaningAppropriate use
real-timeLive spot, updated within secondsCurrent level, intraday reference
15-min delayedExchange-delayed futures settlement or last tradeAnalysis, not execution timing
proxyA tradeable instrument standing in for an unavailable benchmarkDirection and sentiment only
estimate / heuristicOur own calculation or an indicative published figureScreening and prioritisation, never settlement

4 · Derived statistics

All derived figures are computed from two years of daily closes held in our archive. The formulas are stated so results can be reproduced independently.

  • Moving averages: simple, 20-day and 50-day, on daily closes
  • Realised volatility: standard deviation of daily log returns over a 30-day window, annualised by √252, expressed in percent
  • Period change: 30-day (21 trading days), 90-day (63) and 1-year (252), on close-to-close
  • Range position: where the current price sits between the two-year low and high, in percent
  • Spreads and ratios: gold/silver ratio, Brent−WTI differential, corn/wheat ratio — computed on date-matched closes only
  • Outlier guard: a reported daily change beyond ±15% is treated as suspect stale data and suppressed rather than published

5 · Crop-weather layer

Sixteen growing regions are monitored daily. The regions were selected because they materially influence the agricultural markets we cover — not for geographic balance.

VariableDefinitionWhy it matters
Soil moisture, 3–9 cmVolumetric water content of the root zone (m³/m³)Water actually available to the crop, not just what fell
Precipitation, 7-dayAccumulated rainfall (mm)Water supply
Reference ET₀ (FAO-56)Evapotranspiration demand of a reference crop (mm)Water demand — the other half of the balance
Water balance7-day precipitation minus 7-day ET₀ (mm)Whether the crop is gaining or losing ground
Max temperatureDaily maximum, 7-day mean (°C)Heat stress, particularly during flowering and grain fill

6 · The stress heuristic — and its limits

We publish a 0–100 crop-stress score. It is our own construction, and we want you to understand exactly what it is before you use it.

Construction. Three components, weighted:

  • Water balance (50%) — a 7-day deficit scaled to a normal range
  • Root-zone soil moisture (30%) — scored against dry and saturated thresholds
  • Heat (20%) — accumulation above a 32 °C threshold

Bands: 0–25 benign, 25–45 watch, 45–70 elevated, 70–100 severe.

What this score is not. It is not a yield forecast. It has not been statistically validated against realised harvest outcomes, because we have not yet held the data long enough to do so honestly. It does not know the crop's growth stage, the soil type, the cultivar, or whether the field is irrigated — an irrigated field in a dry region will score as stressed when it is not. It is a screening tool: it tells you which of sixteen regions deserves your attention today. Treat any single reading as a prompt to investigate, never as a conclusion.

Our commitment. As the archive matures we will publish a validation study comparing our stress readings against realised yield data, including where the heuristic performed badly. If it does not hold up, we will say so and change it.

7 · Supply and demand matrix

The matrix maps 11 producing regions across 56 commodity markets. Each cell carries a structural view joined to live price and, where relevant, live crop conditions.

  • Production share — approximate share of global mined or grown output, compiled from IEA, USGS, World Bank, IDB, Afreximbank and Australian DISR publications
  • Role — swing (the region sets the marginal global price), major (materially influences balance), marginal (present but not price-setting)
  • Balance — deficit, tight, balanced or surplus
  • Driver — the single constraint we judge most likely to move that cell
These are analytical judgements. Production shares are indicative estimates drawn from published sources of differing vintage and methodology; they are not audited figures and will not reconcile exactly with any single authority. Balance classifications are our assessment, not a measured inventory position. Where a market is a regulatory construct rather than a physical one — carbon allowances, for example — a "share" figure describes the scheme, not physical supply concentration.

8 · Known limitations

We state these plainly because a reader who does not know the gaps cannot judge the work.

LimitationConsequence for the reader
The firm is new. Berube Capital was founded on 26 August 2026.We have no long-run published track record. Our archive begins at founding and compounds from there. Any performance figure we show is labelled as backtested or as live, never blurred.
No freight or shipping data.We cannot yet speak to the cost of moving a cargo, port congestion, or chokepoint delay — a material part of physical commodity economics.
No positioning data.We do not currently ingest exchange positioning (for example CFTC Commitment of Traders), so we cannot tell you how crowded a trade is.
No physical differentials or assessed prices.We publish exchange and spot references, not assessed physical premia by grade and location. For settlement-grade assessments, a price reporting agency remains necessary.
Delayed data for non-metals.Our energy, agricultural and base metal figures lag the market by roughly fifteen minutes. They are unsuitable for execution timing.
The stress score is unvalidated.See section 6. Screening tool, not forecast.
Single-analyst firm.There is no second reviewer on our judgements today. We compensate with transparent method and reproducible calculation, but that is mitigation, not elimination.

9 · Errors and corrections

We will make mistakes. The test of a research firm is what it does next.

  • We correct in place and disclose. A corrected figure carries a note stating what changed and when. We do not silently amend.
  • Material errors are flagged at the top of the affected page or note, not buried.
  • Automated guards. A daily health check verifies that every published payload serves complete data, that our figures are less than eight hours old, and that no page is broken. It runs at 05:00 East Africa Time.
  • Report an error: berubecapital@gmail.com. We would rather hear it from you than not hear it.

10 · Independence

  • We are not a broker, a dealer, or an executing venue. We hold no client funds and take no orders.
  • We accept no payment, data, or consideration from any party in exchange for a particular conclusion.
  • We are not sponsored by, and hold no commercial relationship with, any exchange, producer, or trading house.
  • Our revenue comes from subscriptions, report sales and research mandates — disclosed, and never contingent on the content of a finding.
  • Where the firm or its founder holds a position in an instrument we write about, that interest is disclosed in the note itself.

11 · Disclosures

Research, not advice. Everything Berube Capital LLC publishes is general information and market research. It is not investment advice, not a personal recommendation, and not an offer or solicitation to buy or sell any instrument. It does not take account of your objectives, financial situation or needs. You should obtain your own professional advice before acting.

Risk. Commodity and derivative trading carries substantial risk of loss and is not suitable for every investor. Leverage magnifies both gains and losses.

Performance. Past performance is not indicative of future results. Backtested results are hypothetical, carry the benefit of hindsight, and do not reflect the impact of real slippage, financing or execution constraints. Where we present a backtest we label it as such and state the assumptions.

Data. Our data is supplied "as is" from the third-party sources listed in section 1. We take reasonable care but do not warrant the accuracy, completeness or timeliness of third-party data, and we are not liable for decisions taken in reliance on it.

Attribution. Crop-weather data is provided by Open-Meteo under the Creative Commons Attribution 4.0 licence. Structural production figures are compiled from publications by the International Energy Agency, the United States Geological Survey, the World Bank, the Inter-American Development Bank, Afreximbank and the Australian Department of Industry, Science and Resources.

BERUBE CAPITAL LLC · Nairobi, Kenya · Founded 26 August 2026 · Founder: Reuben Njihia, CPA(K)